Make fragments

The first step of the scene reconstruction system is to create fragments from short RGBD sequences.

Input arguments

The script runs with python run_system.py [config] --make. In [config], ["path_dataset"] should have subfolders image and depth to store the color images and depth images respectively. We assume the color images and the depth images are synchronized and registered. In [config], the optional argument ["path_intrinsic"] specifies the path to a json file that stores the camera intrinsic matrix (See /tutorial/pipelines/rgbd_odometry.ipynb#read-camera-intrinsic for details). If it is not given, the PrimeSense factory setting is used instead.

Register RGBD image pairs

56# examples/python/reconstruction_system/make_fragments.py
57def register_one_rgbd_pair(s, t, color_files, depth_files, intrinsic,
58                           with_opencv, config):
59    source_rgbd_image = read_rgbd_image(color_files[s], depth_files[s], True,
60                                        config)
61    target_rgbd_image = read_rgbd_image(color_files[t], depth_files[t], True,
62                                        config)
63
64    option = o3d.pipelines.odometry.OdometryOption()
65    option.max_depth_diff = config["max_depth_diff"]
66    if abs(s - t) != 1:
67        if with_opencv:
68            success_5pt, odo_init = pose_estimation(source_rgbd_image,
69                                                    target_rgbd_image,
70                                                    intrinsic, False)
71            if success_5pt:
72                [success, trans, info
73                ] = o3d.pipelines.odometry.compute_rgbd_odometry(
74                    source_rgbd_image, target_rgbd_image, intrinsic, odo_init,
75                    o3d.pipelines.odometry.RGBDOdometryJacobianFromHybridTerm(),
76                    option)
77                return [success, trans, info]
78        return [False, np.identity(4), np.identity(6)]
79    else:
80        odo_init = np.identity(4)
81        [success, trans, info] = o3d.pipelines.odometry.compute_rgbd_odometry(
82            source_rgbd_image, target_rgbd_image, intrinsic, odo_init,
83            o3d.pipelines.odometry.RGBDOdometryJacobianFromHybridTerm(), option)
84        return [success, trans, info]

The function reads a pair of RGBD images and registers the source_rgbd_image to the target_rgbd_image. The Open3D function compute_rgbd_odometry is called to align the RGBD images. For adjacent RGBD images, an identity matrix is used as the initialization. For non-adjacent RGBD images, wide baseline matching is used as the initialization. In particular, the function pose_estimation computes OpenCV ORB feature to match sparse features over wide baseline images, then performs 5-point RANSAC to estimate a rough alignment, which is used as the initialization of compute_rgbd_odometry.

Multiway registration

 86# examples/python/reconstruction_system/make_fragments.py
 87def make_posegraph_for_fragment(path_dataset, sid, eid, color_files,
 88                                depth_files, fragment_id, n_fragments,
 89                                intrinsic, with_opencv, config):
 90    o3d.utility.set_verbosity_level(o3d.utility.VerbosityLevel.Error)
 91    pose_graph = o3d.pipelines.registration.PoseGraph()
 92    trans_odometry = np.identity(4)
 93    pose_graph.nodes.append(
 94        o3d.pipelines.registration.PoseGraphNode(trans_odometry))
 95    for s in range(sid, eid):
 96        for t in range(s + 1, eid):
 97            # odometry
 98            if t == s + 1:
 99                print(
100                    "Fragment %03d / %03d :: RGBD matching between frame : %d and %d"
101                    % (fragment_id, n_fragments - 1, s, t))
102                [success, trans,
103                 info] = register_one_rgbd_pair(s, t, color_files, depth_files,
104                                                intrinsic, with_opencv, config)
105                trans_odometry = np.dot(trans, trans_odometry)
106                trans_odometry_inv = np.linalg.inv(trans_odometry)
107                pose_graph.nodes.append(
108                    o3d.pipelines.registration.PoseGraphNode(
109                        trans_odometry_inv))
110                pose_graph.edges.append(
111                    o3d.pipelines.registration.PoseGraphEdge(s - sid,
112                                                             t - sid,
113                                                             trans,
114                                                             info,
115                                                             uncertain=False))
116
117            # keyframe loop closure
118            if s % config['n_keyframes_per_n_frame'] == 0 \
119                    and t % config['n_keyframes_per_n_frame'] == 0:
120                print(
121                    "Fragment %03d / %03d :: RGBD matching between frame : %d and %d"
122                    % (fragment_id, n_fragments - 1, s, t))
123                [success, trans,
124                 info] = register_one_rgbd_pair(s, t, color_files, depth_files,
125                                                intrinsic, with_opencv, config)
126                if success:
127                    pose_graph.edges.append(
128                        o3d.pipelines.registration.PoseGraphEdge(
129                            s - sid, t - sid, trans, info, uncertain=True))
130    o3d.io.write_pose_graph(
131        join(path_dataset, config["template_fragment_posegraph"] % fragment_id),
132        pose_graph)

This script uses the technique demonstrated in /tutorial/pipelines/multiway_registration.ipynb. The function make_posegraph_for_fragment builds a pose graph for multiway registration of all RGBD images in this sequence. Each graph node represents an RGBD image and its pose which transforms the geometry to the global fragment space. For efficiency, only key frames are used.

Once a pose graph is created, multiway registration is performed by calling the function optimize_posegraph_for_fragment.

54# examples/python/reconstruction_system/optimize_posegraph.py
55def optimize_posegraph_for_fragment(path_dataset, fragment_id, config):
56    pose_graph_name = join(path_dataset,
57                           config["template_fragment_posegraph"] % fragment_id)
58    pose_graph_optimized_name = join(
59        path_dataset,
60        config["template_fragment_posegraph_optimized"] % fragment_id)
61    run_posegraph_optimization(pose_graph_name, pose_graph_optimized_name,
62            max_correspondence_distance = config["max_depth_diff"],
63            preference_loop_closure = \
64            config["preference_loop_closure_odometry"])

This function calls global_optimization to estimate poses of the RGBD images.

Make a fragment

134# examples/python/reconstruction_system/make_fragments.py
135def integrate_rgb_frames_for_fragment(color_files, depth_files, fragment_id,
136                                      n_fragments, pose_graph_name, intrinsic,
137                                      config):
138    pose_graph = o3d.io.read_pose_graph(pose_graph_name)
139    volume = o3d.pipelines.integration.ScalableTSDFVolume(
140        voxel_length=config["tsdf_cubic_size"] / 512.0,
141        sdf_trunc=0.04,
142        color_type=o3d.pipelines.integration.TSDFVolumeColorType.RGB8)
143    for i in range(len(pose_graph.nodes)):
144        i_abs = fragment_id * config['n_frames_per_fragment'] + i
145        print(
146            "Fragment %03d / %03d :: integrate rgbd frame %d (%d of %d)." %
147            (fragment_id, n_fragments - 1, i_abs, i + 1, len(pose_graph.nodes)))
148        rgbd = read_rgbd_image(color_files[i_abs], depth_files[i_abs], False,
149                               config)
150        pose = pose_graph.nodes[i].pose
151        volume.integrate(rgbd, intrinsic, np.linalg.inv(pose))
152    mesh = volume.extract_triangle_mesh()
153    mesh.compute_vertex_normals()
154    return mesh

Once the poses are estimates, /tutorial/pipelines/rgbd_integration.ipynb is used to reconstruct a colored fragment from each RGBD sequence.

Batch processing

191# examples/python/reconstruction_system/make_fragments.py
192def run(config):
193
194    print("making fragments from RGBD sequence.")
195    make_clean_folder(join(config["path_dataset"], config["folder_fragment"]))
196
197    [color_files, depth_files] = get_rgbd_file_lists(config["path_dataset"])
198    n_files = len(color_files)
199    n_fragments = int(
200        math.ceil(float(n_files) / config['n_frames_per_fragment']))
201
202    if config["python_multi_threading"] is True:
203        from joblib import Parallel, delayed
204        import multiprocessing
205        import subprocess
206        MAX_THREAD = min(multiprocessing.cpu_count(), n_fragments)
207        Parallel(n_jobs=MAX_THREAD)(delayed(process_single_fragment)(
208            fragment_id, color_files, depth_files, n_files, n_fragments, config)
209                                    for fragment_id in range(n_fragments))
210    else:
211        for fragment_id in range(n_fragments):
212            process_single_fragment(fragment_id, color_files, depth_files,
213                                    n_files, n_fragments, config)

The main function calls each individual function explained above.

Results

Fragment 000 / 013 :: RGBD matching between frame : 0 and 1
Fragment 000 / 013 :: RGBD matching between frame : 0 and 5
Fragment 000 / 013 :: RGBD matching between frame : 0 and 10
Fragment 000 / 013 :: RGBD matching between frame : 0 and 15
Fragment 000 / 013 :: RGBD matching between frame : 0 and 20
:
Fragment 000 / 013 :: RGBD matching between frame : 95 and 96
Fragment 000 / 013 :: RGBD matching between frame : 96 and 97
Fragment 000 / 013 :: RGBD matching between frame : 97 and 98
Fragment 000 / 013 :: RGBD matching between frame : 98 and 99

The following is a log from optimize_posegraph_for_fragment.

[GlobalOptimizationLM] Optimizing PoseGraph having 100 nodes and 195 edges.
Line process weight : 389.309502
[Initial     ] residual : 3.223357e+05, lambda : 1.771814e+02
[Iteration 00] residual : 1.721845e+04, valid edges : 157, time : 0.022 sec.
[Iteration 01] residual : 1.350251e+04, valid edges : 168, time : 0.017 sec.
:
[Iteration 32] residual : 9.779118e+03, valid edges : 179, time : 0.013 sec.
Current_residual - new_residual < 1.000000e-06 * current_residual
[GlobalOptimizationLM] total time : 0.519 sec.
[GlobalOptimizationLM] Optimizing PoseGraph having 100 nodes and 179 edges.
Line process weight : 398.292104
[Initial     ] residual : 5.120047e+03, lambda : 2.565362e+02
[Iteration 00] residual : 5.064539e+03, valid edges : 179, time : 0.014 sec.
[Iteration 01] residual : 5.037665e+03, valid edges : 178, time : 0.015 sec.
:
[Iteration 11] residual : 5.017307e+03, valid edges : 177, time : 0.013 sec.
Current_residual - new_residual < 1.000000e-06 * current_residual
[GlobalOptimizationLM] total time : 0.197 sec.
CompensateReferencePoseGraphNode : reference : 0

The following is a log from integrate_rgb_frames_for_fragment.

Fragment 000 / 013 :: integrate rgbd frame 0 (1 of 100).
Fragment 000 / 013 :: integrate rgbd frame 1 (2 of 100).
Fragment 000 / 013 :: integrate rgbd frame 2 (3 of 100).
:
Fragment 000 / 013 :: integrate rgbd frame 97 (98 of 100).
Fragment 000 / 013 :: integrate rgbd frame 98 (99 of 100).
Fragment 000 / 013 :: integrate rgbd frame 99 (100 of 100).

The following images show some of the fragments made by this script.

../../_images/fragment_0.png ../../_images/fragment_1.png ../../_images/fragment_2.png ../../_images/fragment_3.png